llama : add option to override model tensor buffers (llama/11397)

* llama : add option to override tensor buffers

* ggml : fix possible underflow in ggml_nbytes

diff --git a/ggml/src/ggml.c b/ggml/src/ggml.c
index 161dd3fa94547c9a7b2b83372e65920b5019d65d..3e274d6ae39614b106270de27a4b158c95db072a 100644 (file)
--- a/ggml/src/ggml.c
+++ b/ggml/src/ggml.c
@@ -1159,6 +1159,12 @@ int64_t ggml_nrows(const struct ggml_tensor * tensor) {
 }
 
 size_t ggml_nbytes(const struct ggml_tensor * tensor) {
+    for (int i = 0; i < GGML_MAX_DIMS; ++i) {
+        if (tensor->ne[i] <= 0) {
+            return 0;
+        }
+    }
+
     size_t nbytes;
     const size_t blck_size = ggml_blck_size(tensor->type);
     if (blck_size == 1) {